Abstract
Purpose :
There has been no universal diabetic retinopathy (DR) screening program in Australia. We aimed to investigate the cost-effectiveness of implementing artificial intelligence (AI)-based DR screening in Australian primary care settings.
Methods :
A decision-analytic Markov model was constructed to simulate the progression of DR in a population of 1,262,978 Australian adults with diabetes over a 40-year period. From the healthcare provider’s perspective, we compared the current practice to three potential scenarios employing AI-based screening: (A) substituting the current manual grading, (B) scaling up to the level of patient acceptance, and (C) achieving universal screening. The study outcomes were demonstrated in terms of incremental cost, incremental effectiveness, benefit-cost ratio (BCR), and net monetary benefits (NMB). The analysis utilized a willingness-to-pay (WTP) threshold of AU$ 50,000 per quality-adjusted life year (QALY) and applied a discount rate of 3.5%.
Results :
During 2020-2060, the diabetic cohort with 1,262,978 people is projected to develop 99,957 blindness cases with the status quo. It is projected to cost the healthcare system AU$ 18,021.4 million for DR screening and treatment. Scenario A (replacing manual screening with AI) would prevent 6,709 blindness cases with a gain of 17,187 QALYs over the next 40 years and save the health system AU$ 397.2 million, resulting in a BCR of 15.94 and NMB of AU$ 1,256.5 million. Scenario B (scaling up to patient-acceptance level) would prevent 36,268 blindness cases and save AU$ 1,951.1 million. The BCR would be 15.46 and the NMB would be AU$ 9,680.1 million. If a universal screening program were to be implemented (Scenario C), it would prevent 39,389 blindness cases and save AU$ 2,206.4 million, leading to a BCR of 14.08 and NMB of AU$ 11,257.8 million.
Conclusions :
Implementing AI-based DR screening in primary care is cost-saving compared to the current practice and offers a potential solution to scale up and facilitate DR screening for individuals living with diabetes in Australia.
This abstract was presented at the 2024 ARVO Annual Meeting, held in Seattle, WA, May 5-9, 2024.